Life Activity Modeling of News Event on Twitter Using Energy Function

摘要

This research is the first, exploration on modeling life activity of news event on Twitter. We consider a news event a.s a natural life form, and use an energy function to evaluate its activity. A news event on Twitter becomes more active with a, burst of tweets discussing it. and it fades away with time. These changes of the activity are well captured by the energy function. Then, we incorporate this energy function into the traditional single-pass clustering algorithm, and propose a more adaptive on-line news event detection method. A corpus of tweets which discuss news events was analyzed using our method. Experimental results show that our method not only compares favorably to those of other methods in official TDT measures like precision, recall etc., but also has better time and memory performance, which makes it more suitable for a real system.